The Moshinsky Transform

The Moshinsky transform is the unitary basis change between the coupled two-particle harmonic-oscillator states written in single-particle coordinates and the same states written in center-of-mass plus relative coordinates. It is the trick nuclear shell-model calculations: two-body matrix elements of the nuclear interaction are written naturally in the relative coordinate (where the Yukawa-style nucleon-nucleon force lives), but the shell-model basis is built from single-particle orbitals. The Moshinsky brackets translate.

The whole construction works because of one identity: ${\vec{R}}^2+{\vec{r}}^2={\vec{r}}_1^2+{\vec{r}}_2^2$, where $\vec{R}$ and $\vec{r}$ are appropriately-scaled CM and relative coordinates. That identity is the parallel axis theorem of mechanics in disguise — the same identity that makes Boys' Gaussian product theorem run in quantum chemistry, used differently. See the Gaussian Product Theorem page for the cross-field connection.

The two-body harmonic oscillator

Two nucleons, both of mass $m$, both confined in the same harmonic trap of frequency $\omega$, no nucleon-nucleon interaction yet:

The trap is the harmonic-oscillator approximation to the mean field each nucleon feels from the rest of the nucleus. It's a deliberate fiction — real nuclei don't trap particles in literal SHO potentials — but it's the one fiction that makes everything else tractable, and fixed-with-a-perturbation it's the foundation of the shell model. Eigenstates of this Hamiltonian, before any NN interaction, are products of single-particle harmonic eigenstates: $|n_1{\ell }_1{m}_1⟩|n_2{\ell }_2{m}_2⟩$.

The orthogonal coordinate change

Introduce center-of-mass and relative coordinates in the orthogonal-rotation convention — both scaled by $\sqrt{2}$ so the transformation is a true rotation of the 2-particle configuration space:

This is literally a 45° rotation in the $({\vec{r}}_1,{\vec{r}}_2)$ plane. Orthogonal transformations preserve sums of squares, so

The same rotation applied to the momenta gives $\vec{P}=({\vec{p}}_1+{\vec{p}}_2)/\sqrt{2}$ and $\vec{p}=({\vec{p}}_1-{\vec{p}}_2)/\sqrt{2}$ with ${\vec{P}}^2+{\vec{p}}^2={\vec{p}}_1^2+{\vec{p}}_2^2$. Substitute into $H$ and the Hamiltonian splits cleanly:

Two independent harmonic oscillators, both of mass $m$, both of frequency $\omega$. That same-mass, same-frequency coincidence is what everything that follows — it's why CM and relative motion live on the same oscillator quantum-number grid, with the same energy spacing $\hslash \omega$.

Two bases for the same Hilbert space

The 2-body harmonic Hilbert space now has two natural orthonormal bases:

• Single-particle (coupled): $|n_1{\ell }_1{n}_2{\ell }_2;LM⟩$ — product of two single-particle 3D harmonic eigenstates, angular momenta coupled to total $L$. This is the shell-model basis.
• CM-relative (coupled): $|N\mathcal{L}n\ell ;LM⟩$ — product of a CM 3D harmonic eigenstate $|N\mathcal{L}⟩$ and a relative-coordinate 3D harmonic eigenstate $|n\ell ⟩$, angular momenta coupled to the same total $L$. This is the basis where the NN interaction is diagonal in $\mathcal{L}$.

Both bases diagonalize $H_{\text{harmonic}}$, and both run over the same Hilbert space, so they're related by a unitary transformation. The matrix elements of that transformation are the Moshinsky brackets:

Three crucial properties make these brackets useful in practice:

1. Finite sum. The total oscillator quantum number is conserved: $2n_1+{\ell }_1+2n_2+{\ell }_2=2N+\mathcal{L}+2n+\ell$. So expanding a fixed-shell two-particle state only mixes finitely many CM-relative states.
2. Tabulated. Standard tables (Moshinsky 1959, Brody & Moshinsky 1967) and modern codes give the brackets directly. They're pure linear-algebra coefficients, computable once and reused forever.
3. Locality of the NN interaction. The strong nucleon-nucleon force depends only on $\vec{r}={\vec{r}}_1-{\vec{r}}_2$ (and spin/isospin operators). In the CM-relative basis it's diagonal in the CM quantum numbers $(N,\mathcal{L})$, so the matrix elements collapse to one-dimensional radial integrals in $\vec{r}$.

What this buys for shell-model calculations

A two-body matrix element in the shell-model basis is, in principle, a six-dimensional integral over ${\vec{r}}_1,{\vec{r}}_2$. Direct evaluation against a realistic NN force (Argonne, CD-Bonn, chiral EFT...) would be prohibitive at scale. The Moshinsky transform converts each such matrix element into a finite sum of products:

where the outer brackets are Moshinsky coefficients (tabulated linear algebra) and the inner matrix element is a single-variable radial integral of $V_{NN}$ against harmonic-oscillator radial wavefunctions in $r=|\vec{r}|$. The six-dimensional problem becomes a one-dimensional integral, repeated finitely many times, with the geometry handled by closed-form coefficients. Without this reduction, no-core shell model and configuration-interaction nuclear calculations would not run at any scale.

Why this works only for the harmonic oscillator

The whole construction rests on the fact that ${\vec{r}}_1^2+{\vec{r}}_2^2$ is invariant under the orthogonal rotation that produces $(\vec{R},\vec{r})$. The harmonic-oscillator potential is exactly quadratic in coordinates, so its shape is preserved by that rotation. Any other potential — Coulomb, Yukawa, square-well — does not have this invariance, and its CM-relative decomposition would mix CM and relative coordinates in non-separable ways.

This is the principal axis theorem of linear algebra wearing a shell-model costume: any quadratic form on ${\mathbb{R}}^n$ can be diagonalized by an orthogonal transformation. The 2-body harmonic oscillator IS a quadratic form on the 6-dimensional configuration space, so it diagonalizes. Anharmonic corrections and the NN interaction are added as perturbations on top of this clean separation.

The same identity in quantum chemistry

The orthogonal-rotation identity ${\vec{R}}^2+{\vec{r}}^2={\vec{r}}_1^2+{\vec{r}}_2^2$ appears in quantum chemistry too, in a different guise. Substitute ${\vec{r}}_1\to \vec{r}-\vec{A}$ and ${\vec{r}}_2\to \vec{r}-\vec{B}$ (the orbital centers of two basis functions). The identity becomes

which is the equal-weight case of the parallel axis theorem, and exponentiating it gives Boys' Gaussian product theorem — the algebraic identity that collapses 4-center 2-electron integrals down to one-center integrals plus algebra. Quantum chemistry uses the identity as algebra; nuclear physics uses it as a coordinate change. Same fact, different role. The Gaussian Product Theorem page derives the connection in detail.

Unequal masses and many-body Jacobi coordinates

For two particles of unequal mass $m_1,m_2$, the Jacobi coordinates are $\vec{R}=(m_1{\vec{r}}_1+m_2{\vec{r}}_2)/M$ with $M=m_1+m_2$, and $\vec{r}={\vec{r}}_1-{\vec{r}}_2$. The kinetic energy still separates as ${\vec{P}}^2/2M+{\vec{p}}^2/2\mu$ where $\mu =m_1{m}_2/M$ is the reduced mass. For the harmonic potential, identical-spring trapping $\frac{1}{2}(m_1{\omega }^2{\vec{r}}_1^2+m_2{\omega }^2{\vec{r}}_2^2)$ still separates cleanly, but the CM and relative oscillators now have different effective masses ($M$ and $\mu$ respectively), so the same-shell mixing property is partly lost. Equal masses are the sweet spot that keeps the Moshinsky brackets clean.

For more than two particles, you generalize to many-body Jacobi coordinates, recursively building CM + relative-of-CM coordinates one pair at a time. The same identity holds at each step (always a weighted orthogonal-rotation invariance of the quadratic form), and talmi-Moshinsky brackets for 3, 4, ... particles are the natural generalization. The literature on no-core shell model and ab-initio nuclear structure is largely a story of carrying this machinery to higher particle counts efficiently.

Related on this site

• The Gaussian Product Theorem is the Parallel Axis Theorem — the QC sibling of this identity, with the algebraic derivation and numerical verification of the equal-weight case.
• Nuclear shell model — the broader theoretical framework these brackets live inside.
• Hartree-Fock-Bogoliubov — mean-field theory for paired nucleons, building on the shell-model single-particle basis.
• Two-electron integrals (from scratch) — what the Gaussian product theorem buys in molecular QC, with code.

References

• Moshinsky, M. (1959). Transformation brackets for harmonic oscillator functions. Nuclear Physics, 13(1), 104-116.
• Brody, T. A., & Moshinsky, M. (1967). Tables of Transformation Brackets for Nuclear Shell-Model Calculations. Gordon and Breach.
• Talmi, I. (1952). Nuclear spectroscopy with harmonic oscillator wavefunctions. Helvetica Physica Acta, 25, 185-234.
• Boys, S. F. (1950). Electronic wavefunctions. I. A general method of calculation for the stationary states of any molecular system. Proc. Roy. Soc. London A, 200(1063), 542-554.